I'm still missing another leak:

𓋹 ./python.exe -m test test_module -R :
0:00:00 load avg: 2.89 Run tests sequentially
0:00:00 load avg: 2.89 [1/1] test_module
beginning 9 repetitions
123456789
.........
test_module leaked [1, 1, 1, 1] references, sum=4
test_module failed

== Tests result: FAILURE ==

1 test failed:
    test_module

Total duration: 2.0 sec
Tests result: FAILURE

I'm still missing another leak:

Gotcha! I have it fixed in 3640843

Ok, we are clean:

𓋹 ./python.exe -m test test_module -R :
0:00:00 load avg: 3.33 Run tests sequentially
0:00:00 load avg: 3.33 [1/1] test_module
beginning 9 repetitions
123456789
.........

== Tests result: SUCCESS ==

1 test OK.

Total duration: 2.0 sec
Tests result: SUCCESS

I was about to make my PR ;-)

I was about to make my PR ;-)

Sorry, I was doing this patch and missed your comments on BPO :(

It's hard to see how that code could ever have passed a refleak check! It seems like would always have leaked that dict. Yet I swear I ran one!

@larryhastings If you already have the PR made I can close this one and we can merge yours if you prefer :)

I don't mind the parallel effort, as long as it gets fixed one way or another. I've caused enough headaches and gnashing of teeth this week--let's try and keep this incident to a minimum!

Let's go with yours, the PR is already up and checks are running.

I've caused enough headaches and gnashing of teeth this week--let's try and keep this incident to a minimum!

No problem Larry! This happens quite a lot for everyone, and the important thing is that we have your patch for 3.10 :)

No problem Larry! This happens quite a lot for everyone,

What rot! This never happened during the release process for 3.4 and 3.5! Nobody ever broke the build, ever. I'm clearly a tiresome outlier.